Research On Monocular Vision Inspection Technology For Forming Quality Of The Split-sleeve

The split-sleeve with a characteristic of the open-loop structure is susceptible to deform due to external forces during the contacting measurement.At present,a series of test mandrels are used to examine their sizes of split-sleeves by means of manual off-line inspection.Obviously,this method has the problems of measurement error and slow speed of detection.So,in this paper,the research on the monocular vision detection technology for the forming quality of the split-sleeve has been carried out based on its defects.Therefore,this topic was of importance with engineering application value.In this paper,we selected the widely used and more compact forward lighting as the image detection of the environment according to the practical application scene.The forming quality of the single split-sleeve and the specifications of the multiple split-sleeves were identified by camera calibration,edge extraction and size identification,and then the detection system was designed and developed.The main contents of this paper were as follows:1.Zhang's calibration algorithm was simplified,and the top-view image with certain deviation was converted into the image without deviation by the relation of view transformation.In order to reduce the complexity of both its calibration and its algorithm,and improve the stability of its calibration,the area with small distortion of the image was selected as its detection area.2.For researching on the split-sleeve image based on the forward lighting,the commonly used algorithms of edge extraction were studied and compared.And the algorithm for extracting the outer edge points of the single split-sleeve based on the grayscale value of the image's cross section was proposed in this paper.The position of the outer edge point of the bush was precisely determined by fitting twice.The algorithm combined the local threshold and parallel refinement has been selected for the edge detection algorithm of the multiple split-sleeves,in order to obtain their edge lines with precision of a single pixel wide.3.For measuring the forming quality of single split-sleeve,the Hough transform circle detection algorithm combined with the least squares roundness error was selected,so that the radius and roundness could be obtained.The edge points of Hough transform circle were used for roundness measurement directly,which could avoid the influence of irrelevant edge points and improve the accuracy of roundness measurement.4.For identifying the specification of multiple split-sleeves,the random Hough transform circle detection method was improved in this paper.Firstly,the extracted edges were filtered to remove the irrelevant edges,so that circle detection algorithm could be performed for each edge points set independently.Then the FCM fuzzy clustering algorithm was used to cluster the detected circles' centers,and the number of split-sleeves and the parameters of the circle could be obtained.Finally,the corresponding specification could be inquired through the radius of the table.5.The mixed programming of Matlab and C++ were used to develop the forming quality inspection system of the split-sleeve,in which Matlab can call functions of OpenCV.In this system the advantages of Matlab and C ++ were fully used and the compatibility of the system were improved.